a systematic search for massive black hole binaries in ... - IOPscience

The Astrophysical Journal, 738:20 (9pp), 2011 September 1 C 2011.

doi:10.1088/0004-637X/738/1/20

The American Astronomical Society. All rights reserved. Printed in the U.S.A.

A SYSTEMATIC SEARCH FOR MASSIVE BLACK HOLE BINARIES IN THE SLOAN DIGITAL SKY SURVEY SPECTROSCOPIC SAMPLE P. Tsalmantza1 , R. Decarli1 , M. Dotti2,3 , and David W. Hogg1,4 1

Max-Planck Institut für Astronomie, Königstuhl 17, D-69117 Heidelberg, Germany; [email protected], [email protected] 2 Max-Planck-Institut f¨ ur Astrophysik, Karl-Schwarzschild-Str. 1, D-85748 Garching, Germany 3 Dipartimento di Fisica G. Occhialini, Universit` a degli Studi di Milano Bicocca, Piazza della Scienza 3, 20126 Milano, Italy 4 Center for Cosmology and Particle Physics, Department of Physics, New York University, 4 Washington Place, New York, NY 10003, USA Received 2011 May 11; accepted 2011 June 6; published 2011 August 9

ABSTRACT We present the results of a systematic search for massive black hole binaries in the Sloan Digital Sky Survey (SDSS) spectroscopic database. We focus on bound binaries, under the assumption that one of the black holes is active. In this framework, the broad lines associated with the accreting black hole are expected to show systematic velocity shifts with respect to the narrow lines, which trace the rest frame of the galaxy. For a sample of 54,586 quasars and 3929 galaxies at redshifts 0.1 < z < 1.5, we brute-force model each spectrum as a mixture of two quasars at two different redshifts. The spectral model is a data-driven dimensionality reduction of the SDSS quasar spectra based on a matrix factorization. We identified 32 objects with peculiar spectra. Nine of them can be interpreted as black hole binaries. This doubles the number of known black hole binary candidates. We also report on the discovery of a new class of extreme double-peaked emitters with exceptionally broad and faint Balmer lines. For all the interesting sources, we present detailed analysis of the spectra and discuss possible interpretations. Key words: methods: data analysis – methods: statistical – quasars: emission lines – quasars: general Online-only material: color figures

1. INTRODUCTION

such as the Sloan Digital Sky Survey (SDSS), a large region of the sky can be probed. To date five spectroscopically identified candidates have been presented: J0927+2943 (Komossa et al. 2008; Bogdanovic et al. 2009a; Dotti et al. 2009), J1536+0441 (Boroson & Lauer 2009), J1050+3456 (Shields et al. 2009b), 4C+22.25 (J1000+2233 in this paper; Decarli et al. 2010a), and J0932+0318 (Barrows et al. 2011). Such a small number of objects is marginally compatible with the theoretically predicted number of sub-parsec BHBs at z 0.7 (5–10, given the merger rate and under reasonable assumptions on the binary lifetime and observability; see Volonteri et al. 2009). The spectroscopic approach has an obvious drawback: a peculiar spectrum does not guarantee the presence of a BHB in the nucleus of the host. As an example, an unobscured BHB with both BHs active could resemble the spectrum of a doublepeaked emitter (see, e.g., Eracleous & Halpern 1994), where broad double-peaked lines are emitted because of the almost edge-on, disk-like structure of the BL region of a single BH. A binary with a single accreting BH would show a single shifted BL. If the shift corresponds to a relatively small velocity along the line of sight (4000 km s−1 ), the same signature could be emitted by a remnant of a binary coalescence, recoiling because of anisotropic gravitational wave emission (e.g., Komossa et al. 2008).5 Finally, both the cases can be reproduced by a chance superposition of two AGNs (or an AGN–galaxy superposition) within the angular resolution of the used spectrograph (e.g., Heckman et al. 2009). The simplest way to discriminate between these scenarios and the BHB hypothesis would be to look for a periodic oscillation of the BL shifts around the host galaxy redshift. However, the orbital period of the binary could be too long to be easily observed (Begelman et al. 1980). For these

Massive black hole (BH) pairs are the natural outcome of mergers through the hierarchical formation of galaxies. Examples of unbound BH pairs, with separations of 1 kpc, have been observed, as in the prototypical case of NGC 6240 (Komossa et al. 2003; see Colpi & Dotti 2009 for a recent review). At separations of roughly a few pc the two BHs start experiencing their own gravitational interaction, binding in a BH binary (BHB). Observing BHBs is challenging, since they cannot be spatially resolved in optical and X-ray. The only spatially resolved BHB candidate to date is hosted by the elliptical galaxy 0402+379 (Maness et al. 2004; Rodriguez et al. 2006). The two flat-spectrum radio sources, corresponding to the two components of the candidate BHB, have a projected separation of ≈7 pc. At the distance of 0402+379 (z = 0.055) this corresponds to few milliarcseconds, an angular scale that can be probed only through radio interferometry. A second BHB candidate is the BL Lac OJ287 (see Valtonen et al. 2008, and references therein). It shows a ≈12 yr modulation in its light curve, that has been interpreted as related to the orbital period of a BHB lurking in the nucleus of the active galactic nucleus (AGN). All the other BHB candidates discussed to date have been identified by studying their optical and near-infrared spectra. In a BHB scenario, the broad lines (BLs) emitted by gas bound to each BH may be redshifted or blueshifted with respect to their host galaxy redshift, as a consequence of the Keplerian motion of the binary (Begelman et al. 1980). Furthermore, the BL region of each BH can be perturbed and stripped by the gravitational potential of the companion, resulting in peculiar flux ratios between BLs with different ionization potential (Montuori et al. 2011). This spectroscopic approach does not suffer any angular resolution limitations: actually, the closer (and more massive) the binary is, the more shifted/deformed the BLs are. Thanks to the existence of large spectroscopic surveys,

5

If the galaxy merger is gas rich, the maximum recoil velocity is expected to be 100 km s−1 (Bogdanovic et al. 2007; Dotti et al. 2010; Volonteri et al. 2010; Kesden et al. 2010). In this case, a significant shift between the different sets of lines would not be compatible with a recoiling BH.

1

The Astrophysical Journal, 738:20 (9pp), 2011 September 1

Tsalmantza et al.

reasons, other possible explanations have been proposed for all the BHB candidates discussed in the literature. To overcome the paucity of BHB candidates and the uncertainties related to their interpretation, we depict two ways. Theoretically a better description of the spectrum of BHB is needed, in order to identify other characteristic signatures of binaries. Few attempts have been made to date (see, e.g., Bogdanovic et al. 2008, 2009b; Shen & Loeb 2010; Montuori et al. 2011). Observationally all the BHB candidates lurking in large spectroscopic catalogs must be identified, through a meticulous study of all the possible spectra, in order to allow for follow-up studies on a significant sample of objects. In this paper, we explore the second path, describing the results we obtained from a comprehensive search of BHBs in the SDSS DR7. The code we use, described in Section 2, automatically detects sources with a spectrum consistent with a BHB, a double-peaked emitter, a superposition, or a recoiling BH. We present all the peculiar objects we find in Section 3, where, for each object, we compare our results with previous findings available in the literature. Conclusions are drawn in Section 4. Throughout the paper, we will assume a standard cosmology with H0 = 70 km s−1 Mpc−1 , Ωm = 0.3, and ΩΛ = 0.7.

with the output of PCA when applied to an extended set of training spectra. We then use the resulting set of components to fit each observed spectrum at the redshift provided by SDSS. After that, we repeat the fitting using two sets of components at different redshifts. Here one set is assumed to be at the SDSS redshift, while the second is free to vary over a broad range of z, corresponding to velocity differences up to 30,000 km s−1 . If a second redshift system is present, we expect the fit to significantly improve when we add the second set of components. This procedure has already been proven to successfully identify four of the known BHB candidates (J0927+2943, J1536+0441, J1050+3456, and J1000+2233; see P. Tsalmantza & D. W. Hogg 2011, in preparation) with an estimate of the velocity shifts between the two sets of lines consistent with what obtained in previous studies. 2.2. The Training Sample To train the HMF we use the same training set of quasar spectra that was used in P. Tsalmantza & D. W. Hogg (2011, in preparation). The sample consists of spectra in the redshift range 0.1–1.5. However, since we are mainly interested in detecting shifts between the narrow and the broad emission lines in the QSO spectra, we check if our method is more sensitive in detecting interesting objects when one of the two sets of components used to fit each spectrum is representative of spectra with only narrow emission lines. To define components that include only narrow lines (NLS), we use galaxy spectra for the training of the HMF. The galaxy sample consists of spectra classified spectroscopically as galaxies in SDSS with S/N > 20 and equivalent width of the [O ii] and [O iii] lines larger than 20 (10,856 spectra). All spectra used for the training and the application of the method are derived from the Seventh Data Release (DR7) of SDSS. Pixels with any of the flags: SP_MASK_ FULLREJECT, SP_MASK_BRIGHTSKY, SP_MASK_ NODATA, SP_MASK_NOSKY or pixels that correspond to zero noise were treated as masked. All the spectra and their noise were moved to the rest frame (assuming z = zSDSS ), resulting in spectra with different spectral coverage. However, both HMF and PCA (the output of which was used as an initialization to the HMF) require common wavelengths for all the training spectra. The common wavelength range was defined by the pixels that included information for at least 10 spectra of each type (i.e., galaxies or QSOs). The final wavelength coverage was 1522.299–8352.183 Å for the QSOs and 3044.388–9193.905 Å for the galaxies, corresponding to 7394 and 4801 pixels, respectively. As a last step before the application of the method all spectra were interpolated to common wavelengths selected for each type of source using cubic splines. The selected wavelengths are uniformly distributed in log space as in the case of the original SDSS wavelengths. For the resulting spectra we interpolated linearly the values of the masked pixels. PCA was performed separately for the QSO and the galaxy sample, using a number of spectra equal to the number of pixels selected for each source. The training spectra were first projected into a subspace orthogonal to the mean spectrum of the data set and the flux in each spectral bin was divided with the rms of the noise in that bin, for all the non-masked pixels in the training sample. The PCA results were used as an initialization to the HMF. The method was run for a subset of approximately 1000 spectra of each type for a different number of components and for 16 iterations, until convergence. This test was also performed

2. THE METHOD 2.1. Present Analysis In this study, we perform an automatic and systematic search for BHB candidates in the SDSS catalog, looking for composite spectra of two sources with a velocity difference up to 30,000 km s−1 . More specifically, we look for the presence of two sets of emission lines (one broad and one narrow) with a small separation between them, caused by the Keplerian rotation of one component of the binary system. For this purpose we use the method described in P. Tsalmantza & D. W. Hogg (2011, in preparation). As a first step we extract a small set of components that can sufficiently reconstruct the SDSS QSO spectra by using Heteroscedastic Matrix Factorization (HMF), a bilinear model optimized with probabilistically justified weighted least-squares objective function, analogous to principal components analysis but making use of the observational noise model. The method uses a subset of data at rest-frame wavelengths, described in detail in Section 2.2, as a training set to define the set of basis functions that minimize the scalar χ2 : χ2 ≡

2 N M fij − K k=1 aik gkj i=1 j =1

σij2

K M + [gk − gk(−1) ]2 ,

(1)

k=1 =2

where the first term corresponds to the total χ 2 of the fit of the training data points fij with errors σij , by a set of K components gkj and coefficients aik , over all the N spectra and the M pixels of the training set. The second term corresponds to a smoothing prior or regularization that prefers small pixel-to-pixel variations in each resulting component. The strength of the smoothing is set by the scalar . An iterative procedure is used to minimize χ2 . In each step, we fix gkj and estimate the optimal aik , and then hold the aik fixed and estimate the optimal gkj . The procedure is repeated until the solution converges. We initialize the fitting 2


Tsalmantza et al.

for four different values (1, 3, 10, and 30) of the smoothing scalar . To perform a simple cross validation, the resulting components were also used to fit another subset of 1000 test spectra for each type. Based on these tests, we decided to use 14 and 7 components for the QSOs and galaxies and values of 10 and 1, respectively. Those numbers were also defined by using the spectra of the known BHB candidates and testing the ability of the components to detect them.

2. In all cases, the fitting of a QSO spectrum by the galaxy components will improve significantly when the QSO components are added to the fit. That is due to the fit of the broad emission lines by the second set of components. To make sure that the observed improvements were not only caused by this, we re-perform the fitting of the spectrum by two sets of QSO components, at the redshifts suggested by the Galaxy–QSO fitting. The fits of the spectra selected using the above criteria were visually inspected independently by two of the authors (P.T. and R.D.). The fitting results showed that there is a lot of contamination, e.g., from obvious double-peaked emitters (DPEs), misclassified objects in SDSS, or spectra with wrongly estimated redshifts. By excluding all those cases, the most interesting objects were identified. In Figure 1, we provide an example of the fitting output for one of those sources. The objects selected here are presented in detail in the following section. We should point out that all the objects selected for their interesting features are at redshifts below 0.8. For higher redshifts the [O iii] line is not included in the spectra and the differences in the fitting are detected mainly based on the [O ii] line, which in most of the cases is very faint or not well detected. For this reason no reliable information for the narrow emission lines are included in most of the spectra at higher redshifts.

2.3. Selection Criteria We apply our newly developed fitting scheme to all the 54,586 QSO spectra of SDSS in the redshift range 0.1–1.5. Additionally, since one of the known BHB candidates (4C+22.25; Decarli et al. 2010a) is classified spectroscopically as a galaxy in the SDSS catalog, we also applied the method to objects with redshift from 0.2 to 1.5, that are classified by SDSS as galaxies and have fiber magnitudes that correspond to u − g < 0.8, g − r < 1.5, r − i < 1.0, and g < 21. We note that the majority of these sources (3518 out of 3929) were targeted as quasar candidates by SDSS. For each spectrum the fitting was performed for all the combinations of the extracted QSO and galaxy components: (1) two sets of QSO components for both the spectral components (hereafter QSO–QSO), (2) a set of QSO components for the spectrum at the SDSS redshift and a set of galaxy components at the second redshift (QSO–Galaxy), and (3) the opposite of case (2) (Galaxy–QSO). Out of the 175,545 fitting results we assigned priority based on the following criteria.

3. RESULTS Our selection produced a list of 32 candidates of particular relevance (see Table 1). For each target, we re-analyzed the SDSS spectrum, modeling it with a power law for the QSO continuum emission, a host galaxy template at the redshift of the NLs, and a template of the iron complex, as described in Decarli et al. (2010b) and De Rosa et al. (2011). We fitted the broad components of Mg ii, Hβ, and Hα with two Gaussian functions at the same peak. This fitting approach aims to better constrain the peak wavelength and is not meant to reproduce the line profile in detail. NLs are masked when fitting the broad component; by construction of our sample, there is a velocity offset between BLs and NLs. This simplifies the measurement of the peak wavelengths of the two components. Peak wavelengths are then converted into velocity shifts:

1. The fit significantly improves by adding a set of components at a second redshift, i.e., it corresponds to large χ 2 difference. The threshold was set based on the values extracted for the four known candidates and by visual inspection of the fitting results for various χ 2 difference values. 2. The peaks of the χ 2 difference do not correspond to fits with unphysical properties (e.g., negative emission lines). Negative residuals are common when fitting the spectra with one set of components. Their strength, and therefore their impact in the results, can vary from very weak features usually caused by details in the continuum fitting and the presence of noise, to very strong features caused by poor fitting of emission lines. To exclude the latter cases (but not the former) from our final sample we examine only peaks of the χ 2 differences that correspond to positive differences between the 99.5% and the 0.5% quartiles of the distribution of fluxes per pixel in each fitted spectrum (for each set of components). 3. The peaks of the χ 2 difference correspond to a difference in redshift between the two sources larger than 0.01.

vBL = c

zBL − zNL . 1 + zNL

(2)

Continuum-subtracted velocity plots of all the interesting targets are shown in Figure 2. A rough classification scheme was set according to (1) the magnitude of the velocity shift, in particular when comparing Mg ii and Balmer lines; (2) the presence of strong asymmetries in the line profiles; 3) the occurrence of secondary bumps or peaks; (4) additional information from other emission lines or from the SDSS images. We define five classes of objects, namely, (1) BHB candidates, which are expected to show similar velocity shifts for all the BLs, and a variety of line profiles (e.g., Shen & Loeb 2010). (2) Quasars with asymmetric line profiles, with small (2000 km s−1 ) shifts of BL peaks. These features are observed in some “normal” type-I AGNs, and they are possibly related to asymmetries in the BL region (e.g., Bentz et al. 2010). They may also be associated with a velocitydependent Balmer decrement of broad lines. (3) Double-peaked emitters (DPEs) characterized by symmetric features in line profiles (e.g., a secondary peak in the red wing of the line,

In the case that Galaxy–QSO fitting was performed to the spectra, there was a lot of contamination to our results due to two additional reasons. 1. In most of the cases, the fitting of the narrow emission lines at the SDSS redshift by the galaxy components was resulting to unusual residuals of the BLs, (e.g., in the case that no significant shift was present between the NLs and BLs), which were then fitted very poorly by the second set of components. To exclude these objects from our final sample we also measured the χ 2 value of the fit by the second set of components restricted to an area of 500 Å around the Hα line and 200 Å around the Hβ line. Peaks corresponding to a very large value of χ 2 at the areas of the BLs were excluded. 3


Tsalmantza et al.

χ2 difference 0 1500 3000

MJD=52823, fiberId=572, plateId=1355, z=0.1993, z_2nd=0.2263

0.10

0.15

0.20 2nd redshift

Total Flux

4 3 2 1 0

z_2nd

4 3 2 1 0

z_SDSS

(x10−16)

4 3 2 1 0 4000

5000

6000 7000 wavelength (A)

0.25

8000

0.30

9000

Figure 1. Example of an interesting object (J1539+3333) detected by the method. Top panel: the χ 2 difference between the fitting of the spectrum with 1 and 2 sets of components at different redshifts. The green and blue lines represent the SDSS redshift (zsdss ) and the one with the largest χ 2 difference (zfit ). Second panel: the fitting of the spectrum (black) with both sets of components (red). Third panel: black line, residuals after subtracting the fit components zsdss . The fit with the set of components at zfit is shown in red. Bottom panel: black line: residuals after subtracting the fit at zfit . Red line: the fit with the set of components at zsdss . (A color version of this figure is available in the online journal.)

at the opposite velocity with respect to a blueshifted peak). Another property of DPEs is that different lines (in particular low- and high-ionization lines) may show very different shapes and shifts (Halpern et al. 1996). These properties are usually associated with a disk-like structure of the BL region (Eracleous & Halpern 1994). (4) Extreme double-peaked emitters (see below). (5) Others, i.e., objects with small shifts or poor signalto-noise spectra, preventing us from a clear interpretation, or objects with very peculiar properties, not belonging to any of the aforementioned classes. Note that all the BHB candidates with small velocity shifts could also be recoiling BHs, though they are expected to be rarer than binaries (Dotti et al. 2009; Volonteri et al. 2010).6 In the following, we will not distinguish between these two cases, including both in the “BHB candidates” class. This classification produced nine BHB candidates, including the five known candidates. For the new four sources, other interpretations are also plausible, including an extremely rare case of quasar–galaxy superposition for one of them. Five quasars show very high velocity shifts (5000 km s−1 ) and relatively faint lines. These objects probably represent extreme cases of DPEs (hereafter, they will be referred to as EDPEs). In the following, we discuss the properties of each source individually, reporting our interpretation on the nature of the object. J0012−1022. The Balmer BLs of this zNL = 0.228 quasar show a peak ∼1 700 km s−1 blueshifted with respect to narrow lines. The line profile is clearly asymmetric. Hα and Hβ have identical profiles, with F (Hα) = 2.6F (Hβ). A bump in the red

wing of Balmer lines suggests that this is a strongly asymmetric double-peaked emitter (see also Strateva et al. 2003; Shen et al. 2011), though Shen & Loeb (2010) showed that the line profile of this source can be ascribed to an unequal mass BHB. Classified: asymmetric line profile. J0155−0857. The Hα line of this source shows a small (∼1500 km s−1 ) redshift with respect to the NLs. A slight asymmetry in the line profile is reported. The asymmetry is clearer in the Hβ profile, which peaks at longer wavelengths (Δv ≈ 2200 km s−1 ). Shen et al. (2011) reported a blueshift of ∼600 km s−1 for Hα and ∼2300 km s−1 for Hβ. The relatively small velocity shift and the difference in the profiles of Balmer lines suggest that this is a normal quasar. Classified: others. J0221+0101. Both Hα and Hβ broad components of this zNL = 0.354 quasar show a ∼1 300 km s−1 shift with respect to the NLs. Line profiles are rather boxy with no obvious asymmetry. Shen et al. (2011) report no significant shift for Hα and an exceedingly pronounced shift for Hβ (∼3100 km s−1 ). Classified: others. J0827+2728. The Balmer lines of this quasar show a small (∼900 km s−1 ) redshift with respect to the NLs, which may also be consistent with a strongly asymmetric line profile. Classified: others. J0918+3156. The peculiar properties of this object were first reported by Bonning et al. (2007). The Mg ii and Hβ BLs are redshifted with respect to the NLs. However, the shift is ∼3000 km s−1 for Hβ and only ∼1050 km s−1 for Mg ii (see Bonning et al. 2007; Shen et al. 2011), suggesting that the shift is due neither to a BHB nor a recoil. Classified: double-peaked emitter.

6

Note that these estimates could be affected by our incomplete understanding of the orbital decay of binaries at subparsec scales (see, e.g., Colpi & Dotti 2009; Lodato et al. 2009).

4


Tsalmantza et al.

Table 1 Summary of the Peculiar Objects Found with Our Code Obj. Name (1) J0012−1022 J0155−0857 J0221+0101 J0829+2728 J0918+3156 J0919+1108 J0921+3835 J0927+2943 J0931+3204 J0932+0318 J0936+5331 J0942+0900 J0946+0139 J1000+2233 J1010+3725 J1012+2613 J1027+6050 J1050+3456 J1105+0414 J1117+6741 J1154+0134 J1207+0604 J1211+4647 J1215+4146 J1216+4159 J1328−0129 J1414+1658 J1440+3319 J1536+0441 J1539+3333 J1652+3123 J1714+3327

R.A. (2)

Decl. (3)

MJD (4)

Plate (5)

Fiber (6)

zNL (7)

zBL (8)

Method (9)

Class. (10)

00:12:24.03 01:55:30.02 02:21:13.15 08:29:30.60 09:18:33.82 09:19:30.32 09:21:16.13 09:27:12.65 09:31:39.05 09:32:01.60 09:36:53.85 09:42:15.12 09:46:03.95 10:00:21.80 10:10:34.28 10:12:26.86 10:27:38.54 10:50:41.36 11:05:39.64 11:17:13.91 11:54:49.42 12:07:55.83 12:11:13.97 12:15:22.78 12:16:09.60 13:28:34.15 14:14:42.03 14:40:05.31 15:36:36.22 15:39:08.09 16:52:55.90 17:14:48.51

−10:22:26.3 −08:57:04.0 +01:01:02.9 +27:28:22.7 +31:56:21.2 +11:08:54.0 +38:35:37.6 +29:43:44.1 +32:04:00.2 +03:18:58.7 +53:31:26.9 +09:00:15.8 +01:39:23.7 +22:33:18.6 +37:25:14.8 +26:13:27.3 +60:50:16.5 +34:56:31.4 +04:14:48.2 +67:41:22.7 +01:34:43.6 +06:04:02.8 +46:47:12.0 +41:46:21.0 +41:59:28.4 −01:29:17.6 +16:58:07.2 +33:19:44.5 +04:41:27.0 +33:33:28.0 +31:23:43.8 +33:27:38.3

52141 52168 51869 52932 52990 53050 52731 53389 53386 52254 52281 52757 51989 53737 52993 53757 52375 53431 52356 51942 52051 52376 53116 53120 53120 52426 54523 53498 54567 52823 52790 54591

0651 0665 0406 1267 1592 1740 1214 1939 1941 0568 0768 1305 0480 2298 1426 2347 0772 2025 0581 0491 0515 0842 1449 1450 1450 0911 2758 1646 1836 1355 1343 2973

072 597 374 066 139 399 293 467 553 039 473 281 480 102 110 513 216 603 226 402 099 530 001 141 130 333 014 283 270 572 593 190

0.228 0.165 0.354 0.321 0.452 0.369 0.187 0.713 0.226 0.420 0.228 0.213 0.220 0.419 0.282 0.378 0.332 0.272 0.436 0.248 0.469 0.136 0.294 0.196 0.242 0.151 0.237 0.179 0.389 0.226 0.593 0.181

0.221 0.170 0.364 0.325 0.457 0.372 0.182 0.698 0.226 0.401 0.237 0.168 0.227 0.377 0.276 0.351 0.301 0.258 0.406 0.253 0.450 0.128 0.287 0.206 0.233 0.140 0.242 0.165 0.373 0.199 0.590 0.186

QG,QQ QQ QG,QQ GQ QQ,GQ QQ QQ QG,QQ QQ,GQ QQ QQ GQ QQ GQ QG,QQ QQ QQ QQ,GQ GQ GQ QQ GQ QQ,GQ QQ,GQ GQ QQ,GQ QQ GQ QG,QQ QG,QQ QG,QQ,GQ QQ

A O O O D O A B O B,D A E A E,B O E,B E B E O A,B O O O O O O D D,B O,B O O,B

Notes. Column 1: quasar name; Columns 2 and 3: right ascension and declination (J2000); Column 4: Modified Julian Date of the SDSS observation; Column 5: plate; Column 6: fiber; Column 7: redshift of narrow lines; Column 8: redshift of BLs; Column 9: method used to select the target. QG: Quasar–Galaxy; QQ: quasar–quasar; GQ: Galaxy–quasar (see the text for the method description). Column 10—Classification: A, asymmetric BL profile; B, black hole binary candidate; D, double-peaked emitter; E, extreme double-peaked emitter; O, others (see Section 3).

Shields et al. (2009a). However, later observational follow-ups have disproved the presence of a galaxy cluster in the source field (Decarli et al. 2009). Classified: BHB candidate. J0931+3204. The spectrum of this quasar shows relatively broad and asymmetric narrow lines, which makes the determination of zNL uncertain. The blue wing of Hα shows a bump at ∼3500 km s−1 with respect to the peak of the line. This feature is not observed in Hβ. This object was not reported in previous studies of quasars with peculiar NL profiles (e.g., Wang et al. 2009; Liu et al. 2010b, 2010a). Shen et al. (2011) flagged this source as a DPE candidate. Classified: others. J0932+0318. This source shows a shift of ∼4000 km s−1 in the peak of the Mg ii and the Hβ lines. Shen et al. (2011) reported two different velocity offsets for Hβ (∼3900 km s−1 ) and Mg ii (∼1100 km s−1 ), respectively, but such a difference is not confirmed by our analysis. The Hβ line appears slightly asymmetric, though the line profile is poorly determined. This object has been studied in detail by Barrows et al. (2011), who suggested it to be a DPE from an asymmetric BL region, or a BHB. Classified: double-peaked emitter or BHB candidate. J0936+5331. The Balmer lines of this zNL = 0.228 quasar show a strongly asymmetric profile, with a bright blue peak at 2100 km s−1 (2250 km s−1 in the analysis by Shen et al. 2011)

J0919+1108. The broad emission lines of this source are slightly redshifted (700–1000 km s−1 , depending on the subtraction of the [N ii] lines). The SDSS image of the quasar reveals a complex morphology, probably resulting by a strong gravitational interaction or a merger with a nearby galaxy. Shen et al. (2011) reported a blueshift for the broad component of Hα and a redshift for Hβ, which is not confirmed in our analysis. Classified: others. J0921+3835. This object shows clearly asymmetric Balmer lines, with a peak at ∼1200 km s−1 blueward (consistent with the values reported by Shen et al. 2011). The Hα/Hβ flux ratio is ∼3.3. An unidentified line is observed at 5812 Å, corresponding to a ∼2300 km s−1 shift with respect to the Hβ rest frame. A similar peak is not observed in the Hα profile. Classified: asymmetric line profile. J0927+2943. One of the known BHB candidates (Bogdanovic et al. 2009a; Dotti et al. 2009). This quasar was first reported by Komossa et al. (2008) as a recoiling BH candidate. Three sets of lines are observed, two narrow systems at z = 0.713 and z = 0.699 and a BL system consistent with the second set of NLs. The shift between the two redshifts is ∼2600 km s−1 . This object has been interpreted as a chance superposition in a galaxy cluster by Heckman et al. (2009) and 5


Tsalmantza et al.

Figure 2. Velocity diagrams of Hα (red, solid lines), Hβ (blue, dotted lines), and Mg ii (green, dashed lines) for all our candidates. The flux of Hβ is scaled up to match the one of Hα or Mg ii. (A color version of this figure is available in the online journal.)

with respect to the NLs. The red wing of Hα is 3.1× brighter than Hβ; the blue wing is 4.0× brighter. No peak or bump is observed in the blue wing at 2100 km s−1 . This source was listed in the “DPE Auxiliary sample” by Strateva et al. (2003). Classified: asymmetric line profile.

J0942+0900. This object has been labeled as a galaxy by the SDSS automatic pipeline, presumably because the BLs are too flat and extended. The Balmer lines are ∼38,000 km s−1 broad (the broadest ever reported) and asymmetric (the blue peak is brighter and peaks around ∼−11,500 km s−1 , while the red side 6


Tsalmantza et al. −1

is fainter, flatter, and extends redward of +20,000 km s ). The flux ratio between Hα and Hβ is 2.6, and the lines have similar profile. This object is the most extreme DPE ever discovered in terms of line width. This object may also represent the prototypal to explain the peculiar features of other (less extreme and fainter) cases (e.g., J1000+2233, J1012+2613, etc.). Classified: extreme double-peaked emitter. J0946+0139. The Balmer lines of this zNL = 0.2203 quasar peak at 1550 km s−1 blueward of the NL system (see also Shen et al. 2011). Boroson & Lauer (2010) reported an “anomalous Hβ profile” for this source. The BLs show no obvious second peak, but the blue wings are slightly more extended than the red ones. Also, the flux ratio between Hα and Hβ is larger in the blue wing. Classified: asymmetric line profile. J1000+2233. This source was serendipitously discovered by our group (Decarli et al. 2010a) out of the SDSS database and share some of the properties of J0942+0900. The Balmer lines appear faint and extremely blueshifted (∼8700 km s−1 ). Also the Mg ii line shows a blueshift, but its magnitude is poorly constrained since the line is only partially covered by the SDSS spectrum. This quasar was labeled as a galaxy by the SDSS pipeline. Shen et al. (2011) reported inconsistent velocity offsets for Hβ (4500 km s−1 redward) and Mg ii (2700 km s−1 blueward). Classified: extreme double-peaked emitter or BHB candidate. J1010+3725. This object shows a complex [O iii] profile (both for the 4959 and the 5008 Å emission lines). Two peaks are observed, with a velocity difference of ∼1400 km s−1 . The blue peaks are fainter. Other narrow emission lines (Hβ, Hα, [S ii]) appear normal. This source was labeled as “anomalous [O iii] profile” by Boroson & Lauer (2010). A peculiar [O iii] profile was also reported by Shen et al. (2011). This quasar was not included in previous studies on double-peaked NL objects (e.g., Wang et al. 2009; Liu et al. 2010b, 2010a). Classified: other. J1012+2613. The Balmer lines of this zNL = 0.3783 quasar show similar properties to those of J1000+2233: the BLs peak ∼6000 km s−1 blueward of the NLs. The red wing of Hα and the blue wing of Mg ii are not covered. Classified: extreme doublepeaked emitter or BHB candidate. J1027+6050. This is the only EDPE already known (Strateva et al. 2003). The blue peaks of Hα and Hβ are 7000 km s−1 blueshifted with respect to NLs. The shift is missed in the analysis by Shen et al. (2011), probably because of the faintness of the lines. Classified: extreme double-peaked emitter. J1050+3456. This source was discovered by Shields et al. (2009b) out of the SDSS database. The broad component of Balmer lines is clearly shifted (∼3400 km s−1 blueward; a similar value was found by Shen et al. 2011). No NL is observed at the redshift of the BLs. Classified: BHB candidate. J1105+0414. This quasar was found in the galaxy sample. The broad Hβ line peaks ∼6000 km s−1 blueward of the narrow component. Due to its faintness, the line profile is poorly constrained. The Mg ii line is equally faint, but no obvious shift is observed, supporting the DPE interpretation for this source. Classified: extreme double-peaked emitter. J1117+6741. The Hα broad component of this source appears slightly redshifted with respect to the NLs. The properties of Hβ are difficult to characterize, due to its intrinsic faintness. Classified: other. J1154+0134. The Hβ and Mg ii lines of this zNL = 0.469 quasar have identical profiles, with a peak ∼3500 km s−1 blueward of the expected wavelengths and a rather broad red

wing. The line profiles resemble the one of other sources with asymmetric lines (e.g., J0012−1022), but the magnitude of the shift and the similarity between Hβ and Mg ii support the BHB hypothesis. The low S/N of the SDSS spectrum hinders any conclusion on the nature of this source. Classified: asymmetric line profile or BHB candidate. J1207+0604. The broad component of the Balmer lines of this quasar is rather symmetric but shifted ∼2500 km s−1 blueward with respect to NLs. The flux ratio between Hα and Hβ is ∼4, roughly constant with respect to the line-of-sight velocity. This source was not included in the compilation by Shen et al. (2011). Classified: other. J1211+4647. The bulk of the Balmer line broad components of this source is shifted ∼1700 km s−1 blueward of the NL system (consistent values are reported in Shen et al. 2011). The Hα/Hβ flux ratio is ∼4, constant along the velocity profile. The Hα line shows a bump at ∼6000 km s−1 in the red wing, possibly revealing the DPE-like nature of this source. Such a feature is not clearly observed for Hβ because of the superposition of the [O iii] doublet. Classified: other. J1215+4146. This quasar shows a peculiar Balmer line profile. The bulk of Hα emission arises from a bright bump in the red wing. The blue side of the line may also present a faint wing, the actual presence of which depends on the continuum modeling. At zero order, the Hβ line shows analogous profile. However, the feature in the red wing is ∼11 times fainter than what observed in Hα. The interpretation of this object is unclear. Boroson & Lauer (2010) labeled this source as a “no BL” quasar. Classified: other. J1216+4159. The Hα broad emission of this quasar is clearly blueshifted (∼2300 km s−1 ). The line profile shows no relevant asymmetry. The Hβ broad component is barely detected, its flux being ∼7 times fainter than Hα. Classified: other. J1328−0129. The broad component of the Balmer lines in this zNL = 0.1514 quasar is blueshifted (∼3100 km s−1 ) with respect to NLs. The line profile is boxy, with no significant asymmetry. The Hα/Hβ flux ratio is ∼5, constant over the velocity range. This object was not included in the analysis by Shen et al. (2011). Strateva et al. (2003) and Bian et al. (2007) listed this source as a DPE. Classified: other. J1414+1658. The bulk of the BLs of this quasars is redshifted (∼1200 km s−1 ) with respect to NLs. The red wing is brighter. The Hα/Hβ flux ratio is ∼4 in the blue wing and around 3 in the red wing. This object was labeled as a DPE candidate by Shen et al. (2011). Classified: other. J1440+3319. The Hα line of this source peaks at ∼3700 km s−1 blueward of the NLs and shows an extended red wing. The Hβ line profile is similar. The properties of this quasar are half the way between the objects with asymmetric line profiles (e.g., J1154+0134) and the typical DPEs, though this source has not been included in any previous compilation of DPEs. Classified: double-peaked emitter. J1536+0441. The peculiar properties of this object were first reported by Boroson & Lauer (2009). The BLs show two peaks, one consistent with the rest-frame of the galaxy as set by NLs, the other significantly blueshifted (∼3400 km s−1 ). Boroson & Lauer (2009) proposed the BHB interpretation for this source. However, following observations covering the red wing of Hα revealed the presence of a bump in the line wing (Chornock et al. 2010), a feature commonly observed in DPEs. Classified: double-peaked emitter or BHB candidate. J1539+3333. From the spectroscopic point of view, the properties of this source are similar to those of J0927+2943. The 7


Tsalmantza et al.

spectrum presents three sets of lines at two different redshifts. Broad Balmer lines (driving the redshift estimate by the SDSS pipeline) and faint NLs are detected at z1 = 0.1993. Another set of (brighter) narrow lines is observed at z2 = 0.2263. The corresponding velocity shift is ∼6600 km s−1 . A careful inspection of the SDSS image of this source reveals an extended stellar wing southward of the quasar. If this belongs to the quasar host galaxy, then it would reveal a strongly perturbed morphology. On the other hand, it could be that this is a superposed galaxy. In this case, since zBL < z2 , the galaxy would be in the background of the quasar. This scenario is usually extremely unlikely, given the high velocity differences (Dotti & Ruszkowski 2010). However, the SDSS image reveals the presence of a rich galaxy cluster southwest of the quasar, which may enhance the galaxy density on the sky plain by few orders of magnitudes. Follow-up observations aimed at directly measuring the redshift of the stellar wing are needed to fully understand the nature of this source. Classified: BHB candidate or other. J1652+3123. The Hβ and Mg ii BLs of this zNL = 0.5929 quasar show a small blueshift (∼500 km s−1 ) with respect to NLs. The line profiles are similar and do not show any significant asymmetry. Our shift estimates are consistent with those reported by Shen et al. (2011). Given the small velocity difference, this source is probably a normal quasar (Bonning et al. 2007). Classified: other. J1714+3327. The Balmer lines of this source show a clear redshift (∼1300 km s−1 , consistent with the values reported in Shen et al. 2011). The line profiles are symmetric. The Hα/Hβ flux ratio is around 3. Classified: others or BHB candidate.

ison, only 5 objects out of 138 in Strateva et al. (2003) have a shift of the brighter peak of Hα larger than 5000 km s−1 , and none of them exceed 7000 km s−1 . Note that the “extreme double-peaked emitter” explanation is possible also for one of the BHB candidates already discussed in the literature (J1000+2233; Decarli et al. 2010a). 4. Out of the remaining 13 quasars, 12 show clear yet relatively small (500–3000 km s−1 ) velocity offsets between NLs and BLs, and no obvious asymmetries in the line profiles of the broad components. 5. The case of J1539+3333 is worth of specific discussion. This quasar has few spectral features in common with the BHB candidate J0927+2943, with a set of broad and narrow emission lines shifted with respect to a brighter set of narrow emission lines. However, the velocity shift is extremely high (≈6600 km s−1 ). As a consequence, this object cannot be explained in terms of a recoiling BH, since the maximum kick velocity has been constrained with fully GR simulations to be 4000 km s−1 (Baker et al. 2008; Herrmann et al. 2007; Campanelli et al. 2007; Schnittman & Buonanno 2007; Lousto & Zlochower 2009; van Meter et al. 2010). A natural explanation could be a superposition of an AGN and a galaxy within a galaxy cluster. Such model has been ruled out for J0927+2943 because no galaxy cluster was observable in the field (Decarli et al. 2009). For J1539+3333 a superposition is not ruled out by observations. The redshift of this source is z ≈ 0.2, close to the theoretically estimated peak of superpositions in clusters (Dotti & Ruszkowski 2010). However, the line shift corresponds to a relative velocity between the two galaxies 1.5 times larger than the maximum relative velocity theoretically expected (Dotti & Ruszkowski 2010). A better comprehension of this peculiar object needs a more detailed study of its field, as already performed for J0927+2943 (Decarli et al. 2009) and J1536+0441 (Decarli et al. 2009; Wrobel & Laor 2009).

4. SUMMARY AND DISCUSSION We presented the outcome of our automatic and systematic search for massive BHBs. We have found nine BHB candidates in the SDSS DR7. Of these, five have already been extensively discussed in the literature. The four new candidates are J1012+2613, J1154+0134, J1539+3333, and J1714+3327. For each one of them a BHB is not the only possible explanation: the peculiar spectrum of J1012+2613 can be explained also as an extreme case of double-peaked emitter; J1154+0134 has a too noisy spectrum to exclude other explanations; J1539+3333 may be a rare superposition of a quasar and a galaxy; the small shift between BLs and NLs in J1714+332 (≈1300 km s−1 ) does not necessarily imply the presence of a BHB. A more detailed understanding of the expected spectral features of BHBs and observational follow-ups are needed to confirm or dismiss the BHB hypothesis for all the nine candidates presented here. Our method also automatically detected a number of other interesting objects with peculiar spectral features.

The authors thank C.A.L. Bailer-Jones and F. Walter for useful discussions and support. D.W.H. was partially supported by the NSF (grant AST-0908357) and a research fellowship from the Alexander von Humboldt Foundation. Funding for the SDSS has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Aeronautics and Space Administration, the National Science Foundation, the U.S. Department of Energy, the Japanese Monbukagakusho, and the Max Planck Society. The SDSS Web site is http://www.sdss.org/. The SDSS is managed by the Astrophysical Research Consortium (ARC) for the Participating Institutions. The Participating Institutions are The University of Chicago, Fermilab, the Institute for Advanced Study, the Japan Participation Group, The Johns Hopkins University, the Korean Scientist Group, Los Alamos National Laboratory, the Max-Planck-Institute for Astronomy (MPIA), the MaxPlanck-Institute for Astrophysics (MPA), New Mexico State University, University of Pittsburgh, University of Portsmouth, Princeton University, the United States Naval Observatory, and the University of Washington.

1. Four objects show strong asymmetries in the line profiles, with a peak offset 2000 km s−1 (either redward or blueward) and a longer wing in the opposite velocity range with no secondary peak. 2. Three objects have BL properties analogous to what typically observed in DPEs, even if the secondary peak is not prominent. None of them appeared in the compilation by Strateva et al. (2003). 3. We provide strong evidence of a new class of extreme double-peaked emitters, with very broad (FWHM > 10,000 km s−1 ) and rather faint emission lines. The main peak of these lines shows huge velocity shifts (>5000 km s−1 ) with respect to the NLs. For a compar-

REFERENCES Baker, J. G., Boggs, W. D., Centrella, J., Kelly, B. J., McWilliams, S. T., Miller, M. C., & van Meter, J. R. 2008, ApJ, 682, L29 Barrows, R. S., Lacy, C. H. S., Kennefick, D., Kennefick, J., & Seigar, M. S. 2011, New Astron., 16, 122

8


Tsalmantza et al. Heckman, T. M., Krolik, J. H., Moran, S. M., Schnittman, J., & Gezari, S. 2009, ApJ, 695, 363 Herrmann, F., Hinder, I., Shoemaker, D. M., Laguna, P., & Matzner, R. A. 2007, Phys. Rev. D, 76, 084032 Kesden, M., Sperhake, U., & Berti, E. 2010, ApJ, 715, 1006 Komossa, S., Burwitz, V., Hasinger, G., Predehl, P., Kaastra, J. S., & Ikebe, Y. 2003, ApJ, 582, L15 Komossa, S., Zhou, H., & Lu, H. 2008, ApJ, 678, L81 Liu, X., Greene, J. E., Shen, Y., & Strauss, M. A. 2010a, ApJ, 715, L30 Liu, X., Shen, Y., Strauss, M. A., & Greene, J. E. 2010b, ApJ, 708, 427 Lodato, G., Nayakshin, S., King, A. R., & Pringle, J. E. 2009, MNRAS, 398, L1392 Lousto, C. O., & Zlochower, Y. 2009, Phys. Rev. D, 79, 064018 Maness, H. L., Taylor, G. B., Zavala, R. T., Peck, A. B., & Pollack, L. K. 2004, ApJ, 602, 123 Montuori, C., Dotti, M., Colpi, M., Decarli, R., & Haardt, F. 2011, MNRAS, 412, 26 Rodriguez, C., Taylor, G. B., Zavala, R. T., Pihlstrom, Y. M., & Peck, A. B. 2006, ApJ, 697, 37 Schnittman, J. D., & Buonanno, A. 2007, ApJ, 662, L63 Shen, Y., & Loeb, A. 2010, ApJ, 725, 249 Shen, Y., et al. 2011, ApJS, 194, 45 Shields, G. A., Bonning, E. W., & Salviander, S. 2009a, ApJ, 696, 1367 Shields, G. A., et al. 2009b, ApJ, 707, 936 Strateva, I. V., Strauss, M. A., Hao, L., Schlegel, D. J., Hall, P. B., Gunn, J. E., ˇ 2003, AJ, 126, 1720 Li, L.-X., & Ivezić, Z. Valtonen, M. J., et al. 2008, Nature, 452, 851 van Meter, J. R., Miller, M. C., Baker, J. G., Boggs, W. D., & Kelly, B. J. 2010, ApJ, 719, 1427 Volonteri, M., Gültekin, K., & Dotti, M. 2010, MNRAS, 404, 2143 Volonteri, M., Miller, J. M., & Dotti, M. 2009, ApJ, 703, L86 Wang, J.-M., Chen, Y.-M., Hu, C., Mao, W.-M., Zhang, S., & Bian, W.-H. 2009, ApJ, 705, L76 Wrobel, J. M., & Laor, A. 2009, ApJ, 699, L22

Begelman, M. C., Blandford, R. D., & Rees, M. J. 1980, Nature, 287, 307 Bentz, M. C., et al. 2010, ApJ, 716, 993 Bian, W.-H., Chen, Y.-M., Gu, Q.-S., & Wang, J.-M. 2007, ApJ, 668, 721 Bogdanovic, T., Eracleous, M., & Sigurdsson, S. 2009a, ApJ, 697, 288 Bogdanovic, T., Eracleous, M., & Sigurdsson, S. 2009b, New Astron. Rev., 53, 113 Bogdanovic, T., Reynolds, C. S., & Miller, M. C. 2007, ApJ, 661, L147 Bogdanovic, T., Smith, B. D., Sigurdsson, S., & Eracleous, M. 2008, ApJS, 174, 455 Bonning, E. W., Shields, G. A., & Salviander, S. 2007, ApJ, 666, L13 Boroson, T. A., & Lauer, T. R. 2009, Nature, 458, 53 Boroson, T. A., & Lauer, T. R. 2010, AJ, 140, 390 Campanelli, M., Lousto, C. O., Zlochower, Y., & Merritt, D. 2007, ApJ, 659, L5 Chornock, R., et al. 2010, ApJ, 709, L39 Colpi, M., & Dotti, M. 2009, ASL, in press (arXiv:0906.4339) Decarli, R., Dotti, M., Falomo, R., Treves, A., Colpi, M., Kotilainen, J. K., Montuori, C., & Uslenghi, M. 2009, ApJ, 703, L76 Decarli, R., Dotti, M., Montuori, C., Liimets, T., & Ederoclite, A. 2010a, ApJ, 720, L93 Decarli, R., Falomo, R., Treves, A., Kotilainen, J. K., Labita, M., & Scarpa, R. 2010b, MNRAS, 402, 2441 Decarli, R., Reynolds, M., & Dotti, M. 2009, MNRAS, 397, 458 De Rosa, G., Decarli, R., Walter, F., Fan, X., Jiang, L., Kurk, J., Pasquali, A., & Rix, H. W. 2011, ApJ, in press Dotti, M., Montuori, C., Decarli, R., Volonteri, M., Colpi, M., & Haardt, F. 2009, MNRAS, 398, L73 Dotti, M., & Ruszkowski, M. 2010, ApJ, 713, L37 Dotti, M., Volonteri, M., Perego, A., Colpi, M., Ruszkowski, M., & Haardt, F. 2010, MNRAS, 402, 682 Eracleous, M., & Halpern, J. P. 1994, ApJS, 90, 1 Halpern, J. P., Eracleous, M., Filippenko, A. V., & Chen, K. 1996, ApJ, 464, 704

9